need help with some calculations

[Ecuproject]

hello guys im working on a small bike ecu project

what im trying to do at the moment is trigger the ignition coil on at a desired angle based on some sensor parameters

now the thing that i got stuck for some days is to decode the engine crankshaft wheel correctly.

lets take this example


my bike use a similar wheel its called 12-1 wheel becouse there is one teeth missing out of 12 so in reality there are 11 teeths . bike also use a sensor to count the teeths so it can adjust the mixture and ignition with some calculations.The sensor report 1 if a tooth is on the sensor or 0 if its not so we got something like this pulse

now as we now eatch tooth is 30 deg since 360/12 = 30

and lets say i want to trigger the ignition 45 deg before TDC(when the piston is on top) or 360 - 45 = 315deg.Now to trigger the ignition coil you need to wait 4.3ms(4300uS) to charge the coil before it fires the spark.
Lets take some numbers to work with. I know for example that for one wheel rotation aka 360degs i need 12345uS that means if we do
60000000 / 12345 = 4860 thats our rpm. Now if 12345uS its for 360
then ( ( 360 - 45 ) / 360 ) * ( 12345 - 4300 ) =6501uS is the time to wait before triggering the ignition coil so the spark will be at 315degs as we want it.I need first to find a way to improve it if its possible and second i want to add a "trigger" offset. Trigger offset is for example when the tooth 9 is on sensor and piston is at TDC that means the piston is on top at 270 deg of the wheel

[Salem]

What happened at the 4th sample cycle?

The first three measured 156, 147, 165 pixels (or 18, 17, 19 divisions respectively).
But the 4th measured 226 pixels (27 divisions).

There's no timebase on your picture (and if you're going to save pictures, please use PNG for all machine generated graphics, not JPG).
But 18 divisions for a 12 pulse cycle gives an average of 1.5 divisions per pulse at 4800 RPM.

What happened at the ?
In just 1/4 of a revolution, the speed slowed to just 2400 RPM (a 3 division gap between pulses) and then sped back up again to something close to 4800 RPM.

[Ecuproject]

That pulse was captured when I kicked the bike so its fast at the beginning then it slows down since the engine its not working.Rpm while kicking is arroud 700-1000

[Click_here]

I'd say that a good place to start is to make some code that can pick when the 0deg point is.

It would also need to be able to recover from a faulty reading.


What are you using to read the value of the sensor? Is it an anologue port of a microcontroller? (Watch out for the ignition spark creating noise in the circuitry - Seporating ground, shielding, ect...)


The first place that I'd start is measuring the time between the fall from 1000 (mV?) - From there you can start trying to find the larger gap.

Also, have a look around for what other people have done

[Ecuproject]

well for now im using arduino mega

and im using interrupts to capture the trigger from sensor when it sees a tooth

void TriggerIsr() { //this is the function that gets called by the interrupt
    TriggerTime = micros(); //returns the number of microseconds


    PreviusTriggerDuration = TriggerDuration;
    TriggerDuration = TriggerTime - PreviusTriggerTime;


    if (( PreviusTriggerDuration >  (3 * TriggerDuration) >> 1)  ) {
//missing tooth detected


        if (teethCounter < 11 && (synced == true) ) {


            synced = false;


        }else {


            oneRevolution = !oneRevolution; //this is to detect the strokes
            teethCounter = 1;
            synced = true; //this is to check if we are sycned with the wheel or not
            TimeOn360 = PreviusTriggerDuration * 6; //this is the time for a 360 turn

        }


    }else {

        /*

        */


    }


    teethCounter++;
    oneTeethPreviusLastTime = PreviusTriggerTime;
    PreviusTriggerTime = TriggerTime;
}

then i will use a timer to delay for a specific time to trigger the ignition but this is the hard part let me explain more

to charge the coil you need 4000 uS so for example at 12.000 rpm

the delay of 4000 microseconds is at 288 degres based on this calculation:

4000 * 360 / (60000000 / 12000) = 288 degs

now if i want to trigger 45 degs before TDC i have to do

360 - 45 = 315degs

if my wheel is 12-1 and lets say the tooth 4 is when the piston is at TDC that means (4 - 1) * 30 = 90degs 360 - 90 = 270degs

270 degs is the offset i need to add to the crankshaft angle calculation

now to get the crankshaft angle im doing the following

crankAngle = (teethCounter - 1) * 30 //this will give 30 deg acuracy
crankAngle += 270 //crankshaft offset for TDC

elapsedTime = (micros() - PreviusTriggerTime)
crankAngle += (elapsedTime * 360) / PreviusTriggerTime

if i didnt had the dwell time delay to charge the coil it could be simple be like this
example : crankAngle = 185
rpm = 12000
ignition trigger deg before tdc = 315

315 - crankAngle = 130 degs so (130 / 360) * (60000000 / rpm) =
1805uS

so i just had to delay for 1805 microseconds before triggering the coil.

So my big problem is how to calculate the time for both dwell time (4000 uS) and ignition advance angle(360 - 45 = 315) time so i can set my timer and wait to trigger the spark

[flp1969]

Maybe you can use a Schmitt Trigger, using transistors as signal conditioning, transforming your signal in "square" wave. This will filter noise and give you a "stable" signal to work with.
Why transistors? Because the level of the signal is not sufficient to use 'logical ports' (TTL 'high' is 2.8 V and CMOS circuits is, more or less, 2 V, minimum).

[Ecuproject]

hey i dont have problem with the signal im just having a problem with a calculation check the above post

[Click_here]

45deg is going to be tricky because it doesn't line up with key - The best thing that you can do is an approximation.

The reason is that you can't be sure that the engine is still in steady state.

Have you considered more of a camshaft approach?